Use of Hadoop Framework for Web Based Sentiment Analysis

Current era is of social networking sites, petabytes of data is generated daily on web. Millions of people are posting their likes, dislikes, comments daily on social networking sites. In this system, we are proposing a model that will extract the sentiment from a famous micro blogging site, Twitter, where users post their opinions for everything. Twitter is an online web application which contains lots of data that can be a structured or semi-structured or un-structured data format. We can collect the data from the twitter by using Apache Hadoop(BIG DATA) eco-system using online streaming tool Flume. There are different types of analysis that can be done on the collected data. So here we are taking sentiment analysis, for this we are choose to use Hive and its queries to give the sentiment data based up on the groups that we have defined in the HQL (Hive Query Language).Proposed model uses modified version of Naïve Bayes machine learning algorithm. Our modifications introduce neutral class by eliminating class conditional independence assumption of Naïve Bayes classifier by considering probability intersection between positive and negative classes. Algorithm results are improved by reducing words in tweet to their root form through mechanism of pre-processing before passing them to sentiment analyser. Hence, proposed system classifies tweets as positive, negative or neutral with respect to a query term. This system may prove useful for the enterprises who want to know the feedback about their product brands or the customers who want to improve their productivity and this system may also can be beneficial for election exit polls.